Burkholderia cenocepacia strain J2315 (obtained from the BCCM/LMG Bacteria Collection, Belgium) was used for the construction of transposon mutant pool. The wild type (WT) strain was routinely grown on Ashdown agar and Brain Heart Infusion (BHI) broth at 37°C with agitation (250 rpm). For preparation of electrocompentent cells and recovery of mutants, 2X LB agar supplemented with 500 μg/mL tetracycline were used. B. cenocepacia H111 WT strain and three H111 Tn5 transposon mutants: lon::Tn5, trpA::Tn5, and trpF::Tn5 (Schwager et al., 2013) were used for in vitro growth analysis. The mutants were maintained on LB agar supplemented with 50 μg/mL kanamycin.

To construct the transposon mutants, the EZ-Tn5 <TET-1> Insertion Kit (Cat. No. EZ1921T, Epicentre, Illumina) was used. Transposome was prepared according to the manufacturer’s instructions and kept in -20°C until use. Electrocompetent cells were prepared as described by Dubarry et al. (2010) with minor modifications. The overnight bacteria culture was diluted (1:100) in 50 mL SOB medium supplemented with 0.8% glycine until an OD_{595 nm} of 0.08 was reached. The culture was chilled on ice for 30 min and centrifuged at 4°C at for 10 min at 6,000 ×g. Harvested cells were then washed three times with pre-chilled 10% glycerol and reconstituted in 200 μL of 10% ice-cold glycerol. Electrocompetent cells were immediately used for electroporation. One μL of transposome was added to 50 μL of freshly prepared electrocompetent cells and the mixture was left on ice for 30 min. The cell suspension was transferred to an ice-cold electroporation cuvette (Cat. No. 1652082, Bio-Rad; 2-mm-wide gap) and pulsed using the Gene Pulser (Cat. No. 1652660, Bio-Rad) at 2.5 kV, 25 μF, and 200 Ω. Immediately after electroporation, 950 μL of SOC medium was added and electroporated cells were incubated at 37°C with agitation for 2 h before being spread on 2X LB agar supplemented with 500 μg/mL of tetracycline, followed by incubation at 37°C for 24 h. The number of mutants obtained was counted and colonies were then washed with BHI broth, scraped off from plates and kept in 30% glycerol at -80°C. Electrotransformations were performed in multiple batches until the number of ∼500,000 mutants was reached. Given that B. cenocepacia has a genome size of ∼8 Mb, 500,000 insertions (on the assumption of each mutant is unique) would be sufficient to saturate the genome with one insertion expected every 16 bp. Multiple pools of mutants were then combined to generate a large-scale transposon mutant library or input pool. This input pool was aliquoted and kept in -80°C.

To prevent any low representation of a particular mutant population within the approximately 500,000 individual mutants in the input pool, each individual mutant was represented by at least 20,000 CFU in the initial inoculum. Approximately 1 × 10¹⁰ mutants from the input pool were inoculated into 50 mL fresh Luria-Bertani (LB) broth and grown at 37°C with aeration for 24 h (T₁). Subsequently, 50 μL of this culture (T₁) was inoculated into 50 mL of fresh LB media and grown for a further 24 h (T₂) followed by a second passage (T₃). For M9 minimal medium (supplemented with 0.4% of glucose as carbon source), mutants were grown for 72 h at 37°C (T₁) and 1 mL of this culture was transferred to fresh medium and grown for a further 72 h (T₂) (schematic diagram is shown in Supplementary Materials and Methods Figure S2). For each output pool, 5 mL of bacteria culture was used for genomic DNA extraction.

Genomic DNA of the input pool (in duplicate, each representing the same initial library: Input A and Input B) and output pools was extracted using the MasterPure DNA Purification Kit (Cat. No. MCD85201, Epicentre, Illumina). TraDIS libraries were prepared as described by Langridge et al. (2009) with modifications. Generally, 10 μg of genomic DNA was sheared to 100–1500 bp fragments with a sonicator. The fragmented DNA was end repaired and A-tailed with the NEBNext DNA library preparation kit (Cat. No. E6040S, NEB). Adapters were prepared by annealing two oligos, MP_Ad_a and MP_Ad_b. The adapter was ligated to the A-tailed DNA fragments and 200 ng of the adapter-ligated DNA fragments was PCR amplified over 22 cycles using custom made primers Tra_Fp and Tra_Mp_Rp. PCR amplified libraries were then size selected to between 200 and 500 bp in 1% agarose and purified with a Qiagen gel extraction kit (cat #28704). No specific modifications to the above protocol were made to accommodate the high GC content of the B. cenocepacia genome. Each DNA template in the sequencing libraries consisted of (1) a 3′ transposon end and its adjacent genomic sequence, (2) Illumina-specific sequences for flow cell binding, and (3) a 7-bp barcode for sample multiplexing (Meyer and Kircher, 2010). Libraries were then quantified using both Bioanalyzer and quantitative real time PCR (qPCR) using primers qPCR_P5 and qPCR_P7. Prepped libraries were sequenced on the Illumina HiSeq 2000 platform with the custom sequencing primer Tra_SeqP and index primer Tra_IndP. Sequences of oligos and primers used in this study and PCR cycling conditions are provided in Supplementary Materials and Methods.

Colonies of B. cenocepacia H111 WT and Tn5 mutant strains (lon, trpA, and trpF; Schwager et al., 2013) were inoculated into LB broth and cultures incubated at 37°C for 16 h. Overnight cultures were adjusted to an OD_{595 nm} of 0.5 in LB or M9 minimal medium and 500 μL was transferred to 50 mL fresh media (LB or M9 minimal) and incubated at 37°C with shaking at 250 rpm. At each time point (every 2 h interval), aliquots of bacterial cultures were serially diluted to enumerate the number of live bacteria. For each dilution, 10 μL was dropped onto LB agar (in triplicate) and incubated for 24 h at 37°C. Bacterial CFU for each strain at different time points were then calculated.

Sequence reads from the Illumina FASTQ files were filtered for 10 bases matching the 3′ end of the transposon (TAAGAGACAG) with one mismatch allowed. The transposon tag was removed from the matching reads and the reads were then mapped to the Bcc J2315 reference sequence (EMBL accession numbers AM747720, AM747721, AM747722, and AM747723) using SMALT-0.7.2. The precise transposon insertion site (TIS) was determined using Bio::Tradis¹. Gene essentiality was assessed as previously described (Langridge et al., 2009). Briefly, an insertion index (number of insertion sites divided by gene length) was calculated for each gene. Insertion indexes were observed to fit a bimodal distribution corresponding to two essential and non-essential sets, each of the modes was modeled using a gamma distribution or an exponential distribution to fit genes with no observed insertion sites. Log₂ likelihood ratios (LLR) were calculated between the essential and non-essential models and a gene was classified as essential if it had an LLR of < -2. Similarly, a gene was assigned as non-essential if it had an LLR of >2.

The search against available essential genes available in the database of essential genes (DEG;² version 13.3, updated on January 7, 2016; Luo et al., 2014) was performed using BLASTP with the default parameters provided in DEG. BLAST similarities at protein-protein level that resulted in E-values of 10^-10 or less, were considered matches. B. pseudomallei and B. thailandensis orthologs of B. cenocepacia genes were obtained from the Burkholderia Genome Database³ (Winsor et al., 2008). For growth curves, doubling time (g) was calculated from the exponential phase using the formula: g = t log 2/(log N_t – log N₀) where N₀ = number of CFU at a point during log phase, N_t = number of CFU at a different time point during log phase and t = time interval between N₀ and N_t. The data were expressed as mean ± standard error of the mean (SE) from two independent assays. Statistical analyses were performed using the unpaired, two-tailed Student’s t-test.

Sequence reads were deposited in the database of the European Nucleotide Archive with accession number PRJEB13678 and are accessible via http://www.ebi.ac.uk/ena/data/view/PRJEB13678. The sample accession numbers are ERS1124806 (Input_A), ERS1124807 (Input_B), ERS1124795 (Output LB – T₁), ERS1124796 (Output LB – T₂), ERS1124797 (Output LB – T₃), ERS1124801 (Output M9 – T₁), and ERS1124802 (Output M9 – T₂).

Identification of essential genes by TraDIS requires a large-scale transposon mutant library with appropriate saturation density. To construct transposon insertion mutants, we utilized a transposon mutagenesis system that involves the electroporation of transposome (Tn5 transposon-transposase complex) into B. cenocepacia. It has been reported that genetic manipulation of B. cenocepacia strain J2315 can be laborious due to the difficulty in introducing DNA into the bacteria by both electroporation and conjugal transfer (Dubarry et al., 2010). To enhance the electrocompetency of B. cenocepacia J2315, we adopted the electrocompetent cell preparation protocol described by Dubarry et al. (2010) where bacteria were cultured in medium supplemented with glycine and harvested at a lower density. This approach enabled us to increase the transformation efficiency by up to three fold. Each electrotransformation generated approximately 2,000 mutants and more than 200 electrotransformations were performed. Multiple batches of mutants were combined to achieve a large-scale library of ∼500,000 mutants, henceforth referred to as the initial library or “input pool.”

Next, we used TraDIS to identify the precise TISs for each mutant (Langridge et al., 2009). A custom sequencing primer was designed to amplify the last 10 bp of the transposon extending into the adjacent bacterial genomic sequence. For each sequencing library, we generated an average of 20 million reads with more than 92% of the total reads containing the 10 bp transposon tag. Non-tagged reads were discarded prior to sequence mapping. Over 82% of the tagged reads obtained from the two technical replicates (henceforth referred to as Input A and Input B) were uniquely mapped to the reference genome resulting in approximately 315,000 and 230,000 unique TIS, respectively, across the genome (Table 1). A Spearman’s rho correlation coefficient of 0.9625 indicates low variability in both Input A and Input B, validating the reproducibility of our sequencing data (Figure 1). To integrate these two highly correlated data sets, we combined reads generated from both Input A and Input B to increase the sequencing depth and enhance the degree of confidence in identifying unique TIS. In total, we identified 422,585 unique TIS distributed across the genome and the highest density of insertions (about 275,000 unique TIS) was observed in chromosome 1 at an average of one TIS every 14 bp. Approximately 115,000 insertions were mapped to chromosome 2 with one insertion every 27 bp whilst 24,000 TIS were mapped to chromosome 3 with an average insertion gap of 37 bp and about 6,000 TIS were noted on the plasmid at a frequency of one insertion per 16 bp. This high density of insertions obtained was sufficient to saturate the non-essential genes with Tn5 insertions and also provided the level of genome saturation required to identify essential genes by a negative selection screen.

Disruption of a gene crucial for viability can be detrimental and the loss-of-function mutant is non-viable and most probably not present in the initial mutant library. Therefore, we expect to observe a lower or lack of representation of transposon insertions in regions that contain essential genes. In other words, genes with no or very few transposon insertions are most likely to be essential under normal growth or screening conditions. To discount non-disruptive insertions in the 5′ and 3′ ends, we only analyzed TIS located within the sequence spanning the 5–90% portion of each locus as some essential genes may tolerate disruptions at the extreme 3′ and 5′ ends of the open reading frames (Christen et al., 2011; Le Breton et al., 2015). The insertion index of each gene was calculated by normalizing the number of TIS detected per gene by its respective gene length (5–90%). These insertion indexes were then fitted into a bimodal distribution, generating two peaks corresponding to essential and non-essential gene sets (Supplementary Materials and Methods Figure S1). A log2-likelihood ratio of less than -2 was assumed to determine the essentiality cutoff. Whilst the LLR ratio may be dependent on the depth of sequencing coverage, a gene with an LLR of < -2 (insertion index value is less than the cutoff value) is at least four times more likely to be essential. It should be noted that the existence of paralogous genes may hamper the interpretation of genome-scale mutagenesis experiments, hence considerable care is needed to discriminate essential from non-essential functions (Peterson and Fraser, 2001). In this study, we expected that there would be no reads mapped to genes with multiple copies or to repeat regions because sequence reads mapped to more than one region were discarded prior to analysis of essentiality. A lack of reads would in turn be translated as “no insertion” leading to these genes being most likely misinterpreted as “essential” by our analysis criteria. An example of this is a 57-kb duplication located in chromosome 1 previously described by Holden et al. (2009), which corresponds to 57 genes (BCAL0969 to BCAL1026 and BCAL2901 to BCAL2846). To prevent this misinterpretation, a total of 231 genes for which reads mapped to these genes can also be aligned to other positions (non-unique) were excluded from the essential list, many of which (40%) were mobile elements mainly insertion sequences (ISs) and putative transposases. By utilizing these stringent selection criteria, we obtained a final list of 383 genes predicted to be essential in B. cenocepacia (Supplementary Data Table S1). These “essential genes” are defined as the set of genes required to form colonies on 2X LB agar supplemented with tetracycline.

Of these 383 genes, about 90% are located on chromosome 1 (Figure 2). Chromosome 1 harbors a higher proportion of genes encoding for core cellular functions (Holden et al., 2009) which explains the biased localization of essential genes to this chromosome. There is a high degree of gene conservation within chromosome 1 of different Burkholderia species such as B. pseudomallei, B. mallei, and B. thailandensis (Lin et al., 2008; Holden et al., 2009), suggesting that the essential housekeeping genes of high functional importance are evolutionarily conserved. Nonetheless, the relative lack of essential genes identified in chromosomes 2 and 3 of B. cenocepacia may be due to the different essentiality cutoffs applied to each chromosome to compensate for the uneven distribution of TIS observed (more than 50% of the total TIS found on chromosome 1). As the average gaps between each TIS are found to be larger in other chromosomes, lower essentiality cutoffs limit the number of genes that can be categorized as “essential.”

We searched for homologs among genes deposited in the DEGs (Luo et al., 2014) which contains over 15,000 essential genes identified in more than 30 bacterial strains, and we found 87% of our B. cenocepacia essential candidates (332/383) had homologs in the DEG. To evaluate the functions encoded by the candidate 383 essential genes, the genes were functionally classified based on the COG (Clusters of Orthologous Groups) annotation system. As depicted in Figure 3, the distribution of essential genes in each functional category is typical of that reported for different bacteria (Luo et al., 2014) as well as the synthetic Mycoplasma mycoides JCVI-syn3.0, which has a minimal genome smaller than any free-living cell found in nature (Hutchison et al., 2016). About 30% of the genes encode proteins required for information storage and processing, including proteins involved in replication, transcription and translation such as DNA polymerase subunits (dnaN, dnaQ, dnaX, BCAL1963, dnaE, BCAL2675, and BCAL3371), DNA gyrase subunits (gyrA and gyrB) and ribosomal proteins (BCAL0222-BCAL0259). Similarly, a large subset of genes (34%) was enriched in metabolic and biosynthesis pathways, for example genes encoding ATP synthase subunits (BCAL0330–BCAL0336) and NADH dehydrogenase subunits (BCAL2331–BCAL2344) required for energy production. About 15% of the genes are poorly characterized with no assigned COG function of which many are annotated as hypothetical proteins. The majority of these genes are also present in other Bcc species and do not have homologs in the DEG suggesting that the essentiality of these genes may be specific to Bcc.

Using a bioinformatics approach, Juhas et al. (2012) presented the core genome of the order Burkholderiales consisting of 610 orthologous groups present across 51 Burkholderiales species. These groups correspond to 649 genes likely to be indispensable for survival of B. cenocepacia in its natural environment. Of these candidates, 454 genes have homologs in the DEG. We extended our analysis to compare our sets of essential genes identified by TraDIS with the list of genes predicted by the in silico approach (Juhas et al., 2012). From the total of 383 TraDIS predicted genes, 245 were denoted as essential by computational prediction (Figure 4A), including those for core cellular functions such as replication, transcription, and translation as well as cell wall biosynthesis. Suffice to say, this overlapping group consists of an accurate representation of B. cenocepacia essential genes, although the number constitutes only about 64% of our TraDIS data set. The discrepancy between experimental and computational prediction of essential genes is most likely due to: (1) the stringent analysis parameters applied to our TraDIS data set limited the number of genes which were categorized as essential; (2) species-specific essential genes may not be represented in the core genome of Burkholderiales.

As shown in Figure 4B, arnT, arnB, arnC, arnA1, arnA2, and BCAL1935 located in the arn gene cluster harbor significantly low frequencies of transposon insertions and therefore were categorized as essential in our TraDIS analysis. None of the genes in this cluster was deemed essential by computational prediction. In B. cenocepacia, the arn gene cluster consists of two transcriptional units encoding enzymes for the biosynthesis of aminoarabinose (Ara4N), a molecule involved in lipopolysaccharide (LPS) modification exploited by Gram negative bacteria to adapt to various stressors (Ortega et al., 2007; Hamad et al., 2012). The addition of Ara4N to lipid A is not required for in vitro growth as demonstrated in other bacteria (Kawasaki et al., 2005; Lewenza et al., 2005; Lin et al., 2014) or at least under the laboratory conditions in which previous experiments were conducted. However, it has been shown that disruption of Ara4N biosynthesis in B. cenocepacia results in improper assembly of the outer membrane and confers a lethal phenotype (Ortega et al., 2007). Conditional lethal mutants of arnT and arnB but not the BCAL1928 mutant, exhibited severe growth defects when grown under a non-permissive condition (Ortega et al., 2007). Our TraDIS data supports the essentiality of these genes, as transposon insertions are detectable in the coding region of BCAL1928 but not in arnT and arnB, confirming that arnT and arnB are essential. In addition, homologs of arnT (BPSL1474), arnB (BPSL1472), and arnC (BPSL1474) were deemed essential in B. pseudomallei K96243 by TraDIS prediction (Moule et al., 2014). Similarly, in B. thailandensis, the arn gene cluster (BTH_I2189–BTH_I2195) was predicted as essential by Tn-seq (Baugh et al., 2013).

To support our TraDIS prediction, we searched the literature for B. cenocepacia rhamnose-dependent conditional mutants that are non-viable when grown in glucose (non-permissive condition). Construction of a conditional lethal mutant generally involves the insertion of a rhamnose-inducible promoter upstream of the target gene, which allows transcription to occur only in the presence of rhamnose. If the target gene is essential, the mutant will not grow in the absence of rhamnose. A study by Bloodworth et al. (2013) reported the identification of B. cenocepacia essential operons using conditional growth mutants constructed by rhamnose-inducible promoter replacement of essential genes. About 200,000 B. cenocepacia K56-2 transposon mutants were screened for rhamnose-dependent growth and ∼100 mutants showed less than 35% of WT growth in the absence of rhamnose. By determining the insertion sites of the rhamnose-inducible promoter, which controls the expression of downstream genes in the same operon, 50 unique putative essential operons containing 179 genes were identified. Of these, we found 35 operons that contain at least one TraDIS predicted essential gene and together, this corresponds to a total of 77 genes. As expected, these genes encode enzymes that are involved in core metabolic functions such as energy production (ATP synthase subunits and NADH dehydrogenase subunits) and DNA replication (DNA polymerases). A majority of the remaining essential operons (12/15) that do not contain any TraDIS predicted essential genes are small operons which are made up of not more than two genes. Despite these small operons, none of the genes in operon BCAL2388 - BCAL2395, including purD which encodes a phosphoribosylamine-glycine ligase and hemF which encodes a coproporphyrinogen III oxidase, were deemed essential by our TraDIS prediction. Nonetheless, a total of six genes in this operon have homologs in DEG, suggesting that this operon may be indeed essential in B. cenocepacia. In addition, a number of essential genes (dxs, hemE, infB, gyrB, ubiB, valS, BCAL3369, and murJ) previously validated by a conditional lethal phenotype were also deemed to be essential based on our TraDIS prediction (Cardona et al., 2006; Ortega et al., 2007; Juhas et al., 2012; Mohamed and Valvano, 2014). Hence, we are confident that the TraDIS-based prediction of B. cenocepacia essential genes in this study is highly accurate and reliable.

The essentiality of a gene is also dependent on the specific environment that the bacteria is subjected to Peterson and Fraser (2001). Some genes are crucial for the bacteria to adapt to and survive in the environment it occupies where disruption of those genes would result in reduced growth rates or complete growth arrest in that particular environment. Given that B. cenocepacia is highly versatile and can thrive in diverse environmental niches, to understand how B. cenocepacia adapts to nutrient scarcity, we selected nutrient-depleted medium as a test condition in addition to nutrient-rich Luria-Bertani (LB), a common laboratory medium used in many studies on essential genes. To delineate the subset of genes conditionally essential or critical for growth, we cultured the input mutant pool in nutrient-rich Luria-Bertani (LB) and nutrient-depleted M9 minimal medium. Mutants were grown in LB or M9 over several passages and at each time point, individual output pools were harvested (three time points T₁, T₂, T₃ for LB; two time points T₁ and T₂ for M9 as described in the Section “Materials and Methods”). All output pools were sequenced and a summary of the sequencing results is shown in Table 1.

For every output pool, we assessed the essentiality of each gene using an approach similar to that used for the input pool (Supplementary Data Table S2). In the context of this study, genes categorized as conditionally essential/critical for in vitro growth are those of which insertions resulted in reduced mutant growth, rather than complete growth arrest as defined for the input pool. To determine the consensus sets of genes critical for growth in LB and M9, we integrated the lists of predicted essential genes for each time point using the following criteria: (1) genes found essential in at least two of the three LB pools; (2) genes found essential in either one of the two M9 pools. This resulted in 470 genes in the LB data set and 912 genes in the M9 data set (Supplementary Data Tables S3 and S4) whereby the number of genes is inclusive of the essential genes predicted for the input pool. Three different data sets (initial input pool, LB and M9) were then compared as shown in Figure 5. An overlap of 346 genes were deemed as essential for all conditions and a total of 99 genes were conditionally essential under both LB and M9 growth conditions and we refer to these 99 genes as “general essential” critical for bacterial growth in vitro. Up to 463 genes are required for growth under either one of the conditions tested, which we classify as “condition-specific” (Supplementary Data Table S5). Genomic locations of the “general, LB-specific and M9-specific” essential genes are shown in Figure 2.

A relatively high number of amino acid transport and metabolism genes (71/439) were deemed essential specifically in M9 minimal medium. This observation is consistent with conditionally essential genes identified in other bacteria when grown under nutrient-depleted conditions (Joyce et al., 2006; Lee et al., 2015; Turner et al., 2015). We mapped these genes to KEGG (Kyoto Encyclopedia of Genes and Genomes) pathways and found 34 genes involved in the biosynthesis of amino acids (Table 2). Examples of amino acid biosynthesis pathway-related genes identified from our data set are those involved in the biosynthesis of the aromatic amino acids (schematic shown in Figure 6). Shikimate pathway converts phosphoenol pyruvate and erythrose 4-phosphate to chorismate, a principal precursor of aromatic amino acids, over seven metabolic steps (Herrmann and Weaver, 1999). aroK, aroE, and BCAS0366 which are predicted as conditionally essential genes in our TraDIS analysis, encode components of this pathway. Similarly, deemed conditionally essential are genes involved in the conversion of chorismate to phenylalanine and tyrosine (pheA, BCAL2953, and BCAL3004) as well as tryptophan (trpA, trpC, trpD, and trpF). We suspect that trp mutants may be tryptophan auxotrophs and are unable to grow in a tryptophan-depleted environment. It had been demonstrated in Escherichia coli that ΔtrpA, ΔtrpC, and ΔtrpD strains could not grow on M9-glucose (Patrick et al., 2009). In B. cenocepacia strain H111, trpA and trpB mutants are auxotrophic for tryptophan and supplementation of the growth medium with tryptophan restored bacterial growth (Schwager et al., 2013) suggesting that de novo biosynthesis and transport of amino acids is universally important for bacteria to survive when environmental sources are insufficient.

As proof of principle, we evaluated the ability of three B. cenocepacia H111 Tn5 mutants (trpA::Tn5, trpF::Tn5, and lon::Tn5) to grow in both liquid LB and M9 minimal media as compared to the WT isogenic H111 strain. As the trpA, trpF, and lon mutants were lost from the TraDIS output pools as a result of multiple passages in liquid media, we postulated that disruption of trpA, trpF, and lon do not lead to bacterial cell death, but rather, result in reduced growth rates when the individual mutants are grown in either LB or M9 medium. trpA (BCAM0993) and trpF (BCAM0990), which encode tryptophan synthase alpha chain and N-(5′-phosphoribosyl)anthranilate-isomerase respectively, are predicted to be M9-specific, whilst lon (BCAL1994), which encodes an ATP-dependent protease, is conditionally essential for bacterial in vitro growth in both liquid LB and M9 medium. As shown in Figure 7A, growth rates of all three mutants in LB are comparable to that of the WT strain, with doubling times of 0.76 ± 0.04 h (WT), 1.01 ± 0.29 h (lon::Tn5), 0.69 ± 0.06 h (trpA::Tn5), and 0.85 ± 0.14 h (trpF::Tn5). At early exponential phase, although not significant (p > 0.05), lon::Tn5 mutant showed a slightly slower growth rate. When subjected to M9 minimal medium, trpA::Tn5 and trpF::Tn5 mutants exhibited significantly slower growth (doubling times of 7.10 ± 0.19 h and 4.16 ± 0.81 h respectively) as compared to the WT (doubling time of 1.54 ± 0.10 h; p < 0.05; Figure 7B), thereby validating our TraDIS prediction. This observation is also consistent with previous findings that trpA::Tn5 and trpF::Tn5 mutants could not grow in ABC minimal medium (Schwager et al., 2013). Taken together, our results confirm that trpA and trpF are essential for B. cenocepacia growth in nutrient-depleted medium but not in enriched medium. In addition, the lon::Tn5 mutant showed a much slower growth rate (doubling time of 2.47 ± 0.04 h) when grown in M9 minimal, suggesting that lon is in fact required for B. cenocepacia growth under nutrient limiting condition. Although the lon::Tn5 mutant was predicted to grow slower in liquid LB, we did not observe any significant decrease in the overall growth rate. As the TraDIS screen generally involves a mixed bacterial population and is a competition assay by itself, it is possible that the outcome of the TraDIS prediction may not be similarly reflected in this single-strain experiment. The disruption of lon may not directly lead to significant growth reduction in LB medium, rather, it affects the overall competitive fitness of the resulting mutant in a mixed population.